Skip to content

Experiments on different dataset on how to grow networks during training to learn new image categories.

License

Notifications You must be signed in to change notification settings

vkakerbeck/Progressively-Growing-Networks

Repository files navigation

Progressively Growing Neural Networks

This is code to several experiments on different dataset (MNIST, CIFAR-10 and CIFAR-100) on how to grow networks during training and to learn new image categories. It is the code underlying my master thesis which is also provided here for a better understanding and some additional context. The file TestAccuracies.csv contains all my results and is used in the analysis parts of the code. You can of course replicate this data on your own with the code provided but since this will take several days I provide my results already in TestAccuracies.csv. The full PDF version of my master thesis can be found in the main folder.

Blog Posts

Progressive Learning and Network Growing in TensorFlow: https://towardsdatascience.com/progressive-learning-and-network-growing-in-tensorflow-e41414f304d2

How to Improve Your Network Performance by Using Curriculum Learning: https://towardsdatascience.com/how-to-improve-your-network-performance-by-using-curriculum-learning-3471705efab4

Cite this Work

Kakerbeck, V. (2018). Progressively Growing Neural Networks for Scene Graph Generation from Images (Master thesis, University of Osnabrück, Germany). https://doi.org/10.13140/RG.2.2.26794.44488.

Blogposts: Clay, V. (2018, December 21). Progressive Learning and Network Growing in TensorFlow [Blog post]. Retrieved from https://towardsdatascience.com/progressive-learning-and-network-growing-in-tensorflow-e41414f304d2

Clay, V. (2019, January 7). How to Improve Your Network Performance by Using Curriculum Learning [Blog post]. Retrieved from https://towardsdatascience.com/how-to-improve-your-network-performance-by-using-curriculum-learning-3471705efab4

Dependencies

  • jupyter notebook
  • tensorflow
  • numpy
  • pandas
  • pickle

For plotting and data analysis:

  • matplotlib
  • seaborn
  • augmentation module from https://github.com/aleju/imgaug for CIFAR-100 experiments
  • ptitprince for some of the plots
  • statsmodels & scipy for significance tests and linear regression

About

Experiments on different dataset on how to grow networks during training to learn new image categories.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published